Reading Group Issues in Theoretical Linguistics
Friday, November 30th, 2012
Myriam Cantú Sánchez
Session 24 Reading:
Grinstead, J., Baron, A., Vega-Mendoza, M., DelaMora, J., Cantú-Sánchez, M., Flores-Ávalos, B. (2012) “Tense Marking and Spontaneous Speech Measures in Spanish SLI: A Discriminant Function Analysis” Journal of Speech, Language, and Hearing Research 2012, September 19
1. Abstract / 5. Method2. Tense Marking / 6. Results
3. Tense in Spanish SLI / 7. Discussion
4. Research questions / 8. Conclusion
1. Abstract
Introduction
Specific Language Impairment is a language disorder shown in a remarkable delay in the acquisition of certain aspects of the native language of children who do not have any problem of intelligence, hearing, anatomical structure of the articulatory apparatus or sociability. Approximately 5% of the population suffer this condition, and there is evidence that this impairment is related to a familial genetic antecedent. The search for a valid and reliable diagnostic standard for SLI in different languages has been an important concern for language acquisition investigation.
In this article we test the proposal that the tense deficit that has been demonstrated for children with SLI in some languages is also found in child Spanish and that low performance on tense-related measures can distinguish Spanish-speaking children with SLI from those of normal language development.
Method
· We evaluate evidence from:
a) Existing spontaneous production
b) Elicited production
c) Grammaticality judgment studies of finiteness in Child Spanish
· We measure the relationship of seven spontaneous speech measures with previous receptive and expressive measures of finiteness
· We perform a discriminant function analysis using tense as the target variable to classify monolingual, Spanish-speaking children as SLI children or Typically-developing ones
Results
a) Spontaneous speech measures correlate with the results of previous receptive and expressive measures of child Spanish that show a tense deficit
b) SLI children show statistically lower scores than Typically-developing children on six of seven spontaneous speech measures
c) Tense measures and other spontaneous speech measures are shown to provide fair/good sensitivity and specificity in the classification of children as SLI or TD
Conclusion
Our findings support the contention that the tense marking deficit is a plausible clinical marker of SLI for Spanish-speaking children.
2. Tense Marking
Children normally use and judge as grammatical finite and non-finite verb forms from the time they start using verbs with subjects (around 4;6 of age for English-speaking children). In some cases the verb lacks finiteness morphemes (as in 1 and 2); in other cases the missing finiteness morphemes would appear to have been auxiliary verbs (as in 3 and 4):
(1) Eve 2;0
a. It only write on the pad
b. My finger hurts
(2) Peter 3;3
a. Patsy need a screw
b. This goes in there
(3) Naomi 2;1
Me wearing a curtain
(4) Nina 2;0
I popping balloons
These examples are from the same recording session, illustrating the optionality of finiteness marking (Wexler’s 1994 ‘Optional Infinitive Stage’). In Spanish, there is the assumption that typically developing children are adult-like speakers so early in their development of tense marking that there is no perceptible Optional Infinitive Stage in this language. However, recent studies conclude that Southern Romance child languages do pass through an Optional Infinitive Stage:
· Buesa (2006)
· Liceras, Bel and Perales (2006)
· Davidiak and Grinstead (2004)
· Davidson and Goldrick (2003)
· Clahsen, Aveledo and Roca (2002)
· Radford and Planning-Pacheco (1995)
In Spanish forms like ‘habla’ could be either 3rd singular present verbs or nonfinite bare stem verbs. If we infer that all of these forms are verbs in present tense, children look very adult-like. Interestingly, there is evidence that show that this is not the case: children occasionally produce them with overt non-third person singular subjects, as in the following examples:
(5) Eduardo 3;0
Yo quiere hacerlo
I-nom. want (root + e, theme vowel) do-inf cl-acc-sg-masc
‘I wants to do it”
(6) Carlos 3;3
Yo va a buscar
I-nom go-stem to look for-inf
‘I goes to look for’
(7) Graciela 2;6
Hace esto yo
do(root + e, theme vowel) this I-nom.
‘I does this’
In these examples we have non-third person pronouns used with a third person singular verbs. That could mean that these forms -which do not agree with each other- are possibly not marked for tense, since the same word ending (root + e ‘theme vowel’) expresses tense and agreement. Besides, we know that the overwhelming majority of verbal utterances in Spanish lack overt subjects: we consider that the cumulative effect of many ambiguous 3r. singular present/bare stem forms, being used with null subjects results in an overestimate of child Spanish-speakers’ competence with tense marking.
Multiple elicited production studies show that child Spanish-speakers are delayed in producing tense and agreement marking on both real and nonce words until roughly 4;0 – 5;0:
· Bedore & Leonard 2002
· Bedore & Leonard 2001
· Pérez-Pereira 1989
· Kernan & Blount 1966
In these studies, 3rd singular present verb forms, plausibly bare stems, were the most common error produced by the children. In two more recent works, Vega-Mendoza 2010 and Grinstead, Vega-Mendoza and Goodall 2010, 44 monolingual, 3-6 years old Spanish-speaking children were subjects of a test of acceptance. They had the option of selecting two potential nonfinite forms: the bare stem (canta ‘sing’) and the morphological infinitive (cantar ‘to sing’). The results show that the percentages of correctness correlate with age (r= .679, p<.001). When children were asked to choose between finite and nonfinite verbs occurring with overt subjects, they would choose the nonfinite form 37% of the time, as a function of age. At 5;0 their judgment would reach the 100% of accuracy, which coincides with the age in which English-speaking children appear to have an end to the Optional Infinitive phenomenon established by Rice and Wexler.
3. Tense in Spanish SLI
The studies of tense in Spanish SLI have been carried out analyzing data of children from bilingual and monolingual linguistic environments. The two most influential studies for English-Spanish bilingual environments are Bedore and Leonard 2002 and 2005.
In the study of 2002 the authors compared 15 children diagnosed with SLI (3;11-5;6 years of age, MLUw= 2.88) to a group of 15 age controls and 15 MLUw controls. The children with SLI showed significantly lower scores than age controls for 3rd p. singular and plural for present and past. The great majority of all verbal inflection errors produced in the test by all children consisted of producing a 3rd singular present verb form when a different person, number and tense was called for. This is strong confirmation of our bare stem hypothesis because the attempt forms were overwhelmingly replaced by one form which we believe to be the primary nonfinite form in child Spanish.
In the study of 2005, Bedore and Leonard analyzed spontaneous speech data of the same group of children and observed that the children with SLI are significantly worse than age controls regarding the use of 3rd plural present tense.
For a Catalan-Spanish bilingual environment there are the studies of Sanz-Torrent, Serrat, Andreu and Serra 2008 and Bosch and Serra 1997. Bosch and Serra 1997 studied 24 children. 12 language-impaired (mean age 7;6) and 12 normal developing, control children observed that there were significant differences between the two groups in terms of number marking on 3rd person plural forms; (which the authors attribute to phonological processes).
Sanz-Torrent et al 2008 studied the spontaneous speech of 18 bilingual Catalan-Spanish. Six SLI children, six age control group and six MLUw control group. Data was collected at two points in time: Time 1 when children were 3;9 and Time 2 when children were 4;9. The great majority of the children’s utterances at Time 1 and Time 2 are morphological infinitives of 3rd person singular present forms, and the SLI group produces statistically more 3rd person singular present forms than does the MLUw control group at both times.
Regarding Spanish-speaking monolingual children, two studies, Grinstead et al 2009a and Grinstead et al 2009b analyzed two groups of children from Mexico City. The study of 2009a applied the “Grammaticality Choice Task”[1] to a sample of 27 children. Nine SLI children, 9 age control children and 9 MLUw control group. The results show that SLI children were less proficient than either the MLUw control group (p < .001) and the age control (p < .001).
The study of Grinstead 2009b worked with a larger monolingual group: 42 children in Mexico City. Twenty-one of them were diagnosed with SLI (mean age: 5;6) and 21 were the age control group. Children were presented with an elicited production task. The results also showed that SLI children were significantly worse at producing finite verb forms than were the typically-developing age controls. Interestingly, of the 39 errors produced by the typically-developing children, 23% of them were bare stem forms produced with a non-3rd person singular subject, confirming our hypothesis that the bare stem is a non-agreeing, plausibly untensed form that Spanish-speaking children produce.
4. Research Questions
a) Do spontaneous language measures from Spanish-speaking children who took the elicited production and Grammaticality Choice Task tests correlate with the result of these studies?
b) Can the same spontaneous language measures distinguish Spanish-speaking children with SLI from those without it?
c) Can some combination of these spontaneous language measures and the experimental results alluded to constitute useful discriminant functions capable of identifying Spanish-speaking children with SLI as distinct from typically-developing children with a satisfactory degree of sensitivity and specificity?
Basically, we have shown that the Grammaticality Choice Task and our elicited production tense can distinguish typically-developing children from children with SLI. In this study we would like to know if our spontaneous speech measures can do the same.
5. Method
Participants
Fifty-five monolingual Spanish-speaking from Mexico City participated in this study. Twenty-six of them were diagnosed as SLI (mean age: 5;6) and 29 were typically-developing children (mean age: 5;2). SLI children were diagnosed using conventional exclusive and inclusive criteria:
· Nonverbal IQ was determined using a Spanish translation of the WPPSI (Weschler Preschool and Primary Scale of Intelligence)
· Verbal development was measured using the BELE test (Bateria de Evaluación de la Lengua Española) in which children had to score above 85 out of 100
· Children were also given the ‘Restrepo Family Questionnaire’ to determine if there was any antecedent of language problem in their families
· Children were tested through a phonological screen in which they were asked to repeat 24 nonce words that included segments used in Spanish to represent tense in word-final position, with appropriate stress
· Children proved to have conventional levels of hearing capacities; moreover parental report as well as medical history had to suggest no recent episodes of otitis media with effusion
· Neurological test determined that the children had no neurological damage
· Oral structure and oral motor function were examined and it was assured to have normal function
· Parental report and family history interviews ruled out concerns pertaining to social and physical interactions
The 29 children of the control group were also tested with the phonological screen, standardized language test and it was observed that parents and teachers’ reports suggested no abnormalities in their speech or language
Measures
For each transcript of children spontaneous speech, seven distinct indices of grammatical development were calculated:
1) Mean Length of Utterance calculated in words (MLUw)
2) Mean Length of Utterances calculated in morphemes (MLUm)
3) Mean Length of Utterances of the verb phrase (MLUvp) calculated in words over only the verb phrase
4) Mean Length of terminable unit (MLTU), which only counts clausal utterances
5) Errors per terminable unit
6) Number of different words (NDW)
7) Subordination Index (SUB-I), a ratio of the total number of clauses to the total number of T-units
The calculation of NDW and MLUw was done automatically by the CLAN programs from CHILDES, by MacWhinney, 2000. All other IGD calculations were done by two native Spanish-speakers of the Spanish of Mexico City.
While tense is clearly a grammatical phenomenon situated at the syntax-semantics interface, it equally clearly has a lexical component to it, especially with the less compositional, suppletive and irregular verb forms which are unlikely to be the product of syntactic combination.
Reliability
Spontaneous speech samples were transcribed by native speakers of the Spanish of Mexico City, the same dialect spoken by the children. Each recording session was transcribed by a single transcriber and checked by a second transcriber. Coding for MLUm, MLUvp, MLTU, ETU and SUB-I was also carried out by native speakers of Mexico City Spanish. Agreement between coders ranged from 97,18% to 99,22% with a standard deviation between .891 and 4.280.
Procedures
Spontaneous Production Data:
1) A parent of each child signed US and Mexican institutional review board-approved informed consent documents
2) Children’s spontaneous speech was recorded in sessions of 30 minutes using digital video recorders with built-in microphones
3) The typically developing children were recorded in daycare centers and the SLI children in the Instituto Nacional de Rehabilitación, at the speech and hearing clinic section. In both cases children were asked about their friends, toys, games, movies and TV favorite shows. These data were then analyzed with respect to the 7 IGDs
Correlations with IGDs
1) Our first research question is whether our 7 IGDs from spontaneous speech samples of the children who took the Grammaticality Choice Task tests of tense correlate with results of these studies
2) If they do, the correlations would serve to validate the experimental results of these two tests
Contrast between Two Age-Matched Groups
1) In our previous research, we have shown that the Grammaticality Choice Task and our elicited production tense task can distinguish typically-developing children from SLI children. In this study, we ask if our spontaneous speech measures can do the same
2) To answer to our question, we compare 24 SLI children (mean age= 5;5) and 24 age-matched typically-developing children (mean age= 5;5).